Public-health applications in remote sensing

Remote sensing of the environment is critical to warn of imminent, life-threatening dust storms and to reduce risk of exposure to desert dust and hitchhiking bacteria, molds, heavy metals, and other human-health concerns. Desert dust affects cardiovascular and respiratory illness. Respirable particulates result in health-care costs exceeding $11.5 billion annually, with an additional $4.6 billion for lost productivity.¹

As part of the NASA-sponsored Public Health Applications in Remote Sensing (PHAiRS) project, data from NASA's Terra satellite are assimilated into a numerical-dynamical model of dust generation and entrainment, DREAM (Dust Regional Atmospheric Model). Originally developed for the Mediterranean region² and modified in PHAiRS for the southwestern United States,³ DREAM is driven by operational weather-forecast models of the US National Weather Service (the Nonhydrostatic Mesoscale Model, NMM, and the operational model it replaces, eta). The DREAM system simulates and predicts—up to three days in advance—the onset of dust storms and the 3D size-concentration characteristics of the resulting airborne-dust clouds.

Current barren-ground input to the model consists of MOD12 classification (an International Geosphere-Biosphere Programme ecosystem descriptor of land-cover type and dynamics) condensed to a two-class product: bare ground=1 and all other classes=0. Using Moderate Resolution Imaging Spectroradiometer (MODIS) products to replace the bare-ground class from the Olson World Ecosystem Land Cover map—the model's original design parameter (see Figure 1)—improved DREAM's performance significantly (see Figure 2).^4,5

Figure 1. (left) Olson World Ecosystem ‘barren’ category. (right) MOD12 (land-cover type and dynamics)-classified ‘barren’ category.

Figure 2. Dust near-surface concentrations over Odessa, TX, on 13 December 2003, simulated by DREAM (solid lines) using (left) Olsen Land Surface and (right) NASA Moderate Resolution Imaging Spectroradiometer land characteristics compared with measured concentrations (dotted line). Note the left- and right-panel scale differences.

Working with New Mexico's Environmental Public Health Tracking System, we developed a Web-based client server (see Figure 3) to assist asthma and myocardial-infarction surveillance across the state. Timely forecasts will also allow environmental alerts and prepare health-care providers for increases in both inpatient and outpatient services.

Figure 3. The Public Health Applications in Remote Sensing client website⁶allows user options for obtaining model-generated data of dust events, including 72-hour animated surface-dust-concentration forecasts and, post-event, comparison with in situ measurements if available. UTEP: University of Texas at El Paso. AQI: Air-quality index. EPA: Environmental Protection Agency. UTC: Coordinated universal time. C12/C125/C151: UTEP monitor. pm10: Particulate-matter size of 10μm.

Figure 4. Hindcast of 18 July 2007 dust storm across Arizona in the early-morning hours in three time steps. Near-surface AQI contributed by desert dust in color gradations from light blue (AQI 1: 1–50mg/m³)to red (AQI 6: >300mg/m³).

For model verification, we rely on a regionally sparse network of in situ sampling stations to compare model-generated concentrations of particulate matter with sizes of 10 and 2.5μm (PM₁₀ and PM_2.5, respectively). These networks are concentrated in densely populated urban areas, subject to PM₁₀ and PM_2.5 sources generated by human activity. Few particle-sampling sites are available that identify natural versus man-made sources. Future work will address these problems, distinguishing seasonally changing surface characteristics and turning to lidar (light detection and ranging) networks and the A-Train's (NASA's ‘Afternoon’ satellite constellation) CALIPSO (Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observation) and Glory satellites for measurements of aerosol profiles with which to compare model output. To improve the accuracy and usefulness of simulations and forecasts, DREAM-NMM has been tested at 3–5km spatial resolution (see Figure 4). It is currently being installed on the University of Arizona's supercomputer, which will yield greater detail and a wider spatial domain.

Products specifically designed with the end user in mind are being evaluated in key Arizona and New Mexico state offices with operational health and air-quality responsibilities. These products will be modified as needed for public-health services.

This work is sponsored by NASA's Science Mission Directorate, Applied Sciences and Public Health Programs. We acknowledge other members of the PHAiRS team, especially Karl Benedict and Len Flowers, and former members Dazhong Yin and Patrick Shaw, and early contributions and advice by Slobodan Nickovic.